Glioma, a highly aggressive brain cancer, thrives in an immunosuppressive tumor microenvironment (TME). The Unfolded Protein Response (UPR) is a key adaptive stress pathway implicated in therapy resistance. To discover clinically actionable biomarkers, we employed an integrated multi-omics strategy, combining single-cell and bulk transcriptomic data from multiple glioma cohorts. Using hdWGCNA network analysis, we identified a UPR-associated gene module that defined two molecular subtypes with stark survival differences. Through a rigorous machine learning pipeline (Random Survival Forest, LASSO, and CoxBoost), ANXA5 emerged as the central prognostic gene. High ANXA5 expression consistently predicted poorer overall survival across eight independent datasets, validating its robust prognostic value. Functionally, ANXA5 was linked to extracellular matrix remodeling and immune modulation. Multi-omics profiling revealed that ANXA5-high gliomas exhibit a T-cell-inflamed yet immunosuppressive TME, characterized by elevated immune checkpoint expression. Crucially, ANXA5 demonstrated strong predictive power for response to immune checkpoint blockade (ICB), showing significant correlation with nine established immunotherapy response signatures and accurately discriminating responders from non-responders in six independent ICB-treated clinical cohorts (AUC: 0.65-0.78). Genomic analysis associated ANXA5 expression with distinct mutation patterns (EGFR/PTEN vs. IDH1/TP53). In vitro knockdown of ANXA5 confirmed its oncogenic role, as it suppressed glioma cell proliferation and invasion. Our study establishes ANXA5 as a prime example of a translatable biomarker discovered through multi-omics integration. It functions dually as a prognostic indicator and a predictive biomarker for immunotherapy, offering a tangible framework for patient stratification and personalized therapeutic strategies in glioma, thereby bridging a critical gap toward clinical translation.
Wang et al. (Sun,) studied this question.